Optimization and control using response surface methods

Many processes exhibit an optimum that moves in time. Most optimization algorithms assume that the optimum is like a static hill and their procedures rely on this. This paper extends the ideas of Response Surface Methodology (RSM) to allow tracking of dynamic optima. The new algorithm is called DRSM. Many model based control algorithms are structured as optimization problems, albeit for linear processes. DRSM is then also used to develop a model predictive controller that can be used for non-linear, time-varying processes, without the need for a detailed process model. Simulations are used to demonstrate the algorithm on problems in optimization and control. Copyright (C) 1996 Elsevier Science Ltd